AN ABSTRACT OF THE DISSERTATION OF. Rebecca Cullion for the degree of Doctor of Philosophy in. Mechanical Engineering presented on December 05, 2005.

Transcription

1 AN ABSTRACT OF THE DISSERTATION OF Rebecca Cullion for the degree of Doctor of Philosophy in Mechanical Engineering presented on December 05, Title: Void Fraction Variations in a Fractal-Like Branching Channel Network Redacted for privacy Abstract approved: Deborah Pence Two-phase flow in a fractal-like branching microchannel heat sink may have enhanced heat transfer capabilities over single phase flow in the same branching channel network. In order to exploit this potential, a complete understanding of flow boiling in this geometry is required. The fractal-like geometry is similar to that of the human circulatory system; a larger diameter channel, or artery, branches into two channels, or arteries, of smaller diameter. An experimental study of flow boiling in a fractal-like branching channel heat sink was performed. The fractal-like network, fabricated for use in this study, utilized a circular configuration, in which fluid entered the center of the device and flowed radially outward. The resultant, bifurcating pattern is perfectly symmetrical, and has four different branching levels. The channels of the branches range in hydraulic diameter from 218 micron at the inlet to 120 micron at the periphery of the heat sink. High-speed, high-resolution imaging was used to visualize flow regimes and quantify void fraction variations in the channels. Regional and local changes in void fraction were analyzed for the effects of this novel geometry on flow boiling characteristics. Of particular interest was the interaction between different branching levels and the impact of bifurcations on vapor flow.

2 Global measurements of pressure, temperature, flow rate, and power input were also made. Operating conditions included an inlet temperature of 88 C, inlet mass flow rates between 45 and 65 g/min, and two power levels of 61 W and 66 W. Flow regimes observed during qualitative analysis of a base case, defined as q = 66 W, th = 45 g/min, and T1 = 88 C, included single phase, bubbly, and slug flow. Quantitative analysis of the base case indicate that the levels of the fractallike device interact; vapor in a downstream channel affects the flow behavior in an associated upstream channel. In addition, bifurcations associated with the fractallike geometry affects flow characteristics. Bifurcations split downstream vapor flow and redirected vapor flowing upstream. Comparison of base case results with two other cases indicate increased flow rate, and decreased power, independently t produced void fraction variations that followed the same trends as those found in the base case but with different magnitudes and frequencies.

3 Copyright by Rebecca Cullion December 05, 2005 All Rights Reserved

4 Void Fraction Variations in a Fractal-Like Branching Channel Network by Rebecca Cullion A DISSERTATION submitted to Oregon State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Presented December 05, 2005 Commencement June 2006

5 Doctor of Philosophy dissertation of Rebecca Cullion presented on December 05, 2005 APPROVED: Redacted for privacy Major Profeor, representing Mechanical Engineering Redacted for privacy Head of the Department of Mechanical Engineering Redacted for privacy Dean of the/drki.te School I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my disseration to any reader upon request. Redacted for privacy Rebecca Cullion, Author

6 ACKNOWLEDGMENTS I would like to take this opportunity to acknowledge the individuals who made contributions to this work. First, I would like to thank Dr. Deborah Pence for her role as my advisor. Thank you for teaching my undergraduate Thermodynamics class in such a manner that I was inspired to become a professor. Thank you to Dr. Charles Brunner, Dr. Murty Kanury, Dr. James Liburdy, and Dr. Vinod Narayanan for serving on my committee. Special thanks to Dr. Liburdy and Dr. Narayanan for the use of equipment from their research labs. Two students contributed a great deal to this project, Greg Mouchka and Jaime June11. My thanks to both of them for all of their help and the laughs. I cannot finish without acknowledging truffles, coffee cards, gelato certificates, and twenty dollar bills. Thank you so very much Mom, Dad, Chris, Nan, Cheryl, and Jim. Your thoughtfulness brightened every day. Finally, I would like to thank Kyle for making everything better always.

7 TABLE OF CONTENTS Page 1 BACKGROUND LITERATURE REVIEW AND OBJECTIVES Literature Review Microscale Single Phase Flow Macroscale Two-Phase Flow Microscale Two-Phase Flow Research Objectives TEST DEVICE DESIGN AND MANUFACTURING Test Device Design Test Device Manufacturing EXPERIMENTAL FACILITY AND TEST PLAN Manifold Flow Loop and Instrumentation Flow Loop Flow Visualization Data Acquisition Test Plan Procedure DATA REDUCTION AND ANALYSIS Data Reduction Global Data Reduction Regional and Local Data Reduction... 28

8 TABLE OF CONTENTS (Continued) Page Uncertainty Data Analysis RESULTS Qualitative results Quantitative Results Base-Case Regional Quantitative Results Case 1 Regional Quantitative Results Case 2 Regional Quantitative Results Base-Case Local Quantitative Results Case 1 Local Quantitative Results Case 2 Local Quantitative Results CONCLUSION BIBLIOGRAPHY APPENDICES APPENDIX A Calibration APPENDIX B MATLAB Algorithms APPENDIX C Instrumentation

9 Figure LIST OF FIGURES Page 1.1 Fractal-like branching channel network Tree of a fractal-like branching channel network Cross-sectional view of an ideal test device Lengths are radial distances Nichrome heaters located on bottom side of test device Manifold Flow ioop Flow visualization loop Rhodamine 6G absorption and fluorescence emission spectra from http: //probes.invitrogen. corn Regions of interest located in the k = 2 level Definitions associated with the local study Definitions and key terms used when discussing results Four images seconds apart from a base-case movie that demonstrate the observed bubbles, slugs, and vapor fronts Inlet temperature, pressure, mass flow rate, and power input measurements sampled at 1000 Hz, the mean of which was recorded every seven seconds for the thirty minute period Time-averaged void fraction in the k2ar and k2br regions determined for each of the ten movies Regional, time series void fraction variations in the k 2 level of tree 2, associated with t = 14 mm movie Regional, time series void fraction variations, in the k = 2 level of tree 2, associated with t = 14 mm movie between t = 1.5 and 2.5 s Regional, time series void fraction variations, in the k = 2 level of tree 2, associated with t = 3 mm movie Regional, time series void fraction variations, in the k = 2 and k = 3 levels of tree 2, associated with t = 14 mm movie... 43

10 Figure LIST OF FIGURES (Continued) Page 6.9 Time-averaged void fraction of the k = 3 level regions of tree 2 determined from each of the ten movies Regional, time series void fraction in the k = 3 level of tree 2 associated with t = 5 mm movie Regional, time series void fraction in the k 3 level of tree 2 associated with t 5 mm movie when the average void fraction is large and the instantaneous void fraction oscillating Void fraction variations in tree 2 during the t = 20 mm movie. (a) Interaction between k = 3 and k 2 levels. (b) No interaction on the k = 2 level (c) Similar behavior in the k 3 level regions Regional, time series void fraction variations, in the k = 3 level of tree 2, associated with t = 3 mm movie Time-averaged void fraction of k = 2 level regions of both tree 1 and 2 determined from each of the ten movies Regional, time series void fraction in the k = 2 level of trees 1 and 2 associated with t = 28 mm movie Time-averaged void fraction of k = 3 level regions of both tree 1 and 2 determined from each of the ten movies Regional, time series void fraction in the k = 3 level of trees 1 and 2 associated with t = 28 mm movie Inlet temperature, pressure, mass flow rate, and power input measurements of case 1, sampled at 1000 Hz the mean of which was recorded every seven seconds Time-averaged void fraction in the k2ar and k2br regions was determined for each of the ten movies associated with the base case and case Time-averaged void fractions in the k3br and k3cr regions were determined for each of the ten movies associated with the base case and case Comparison of the instantaneous void fraction in tree 2 during the base case t = 15.5 mm movie and case 1 t = 14 mm movie. (a) k 2 level (b) k 3 level... 60

11 Figure LIST OF FIGURES (Continued) Page 6.22 Close-up of the instantaneous void fraction in tree 2 during the t = 15.5 mm movie of case 1 and the t = 14 mm movie of the base case. (a) Case 1 (b) Base Case Comparison of the instantaneous void fraction in tree 2 during the base case, t = 28 mm movie and case 1, t = 28 mm movie. (a) k 2 level (b) k = 3 level Regional, time series void fraction variations in the k = 3 level of tree 2 associated with t = 0 mm movie of the case 1 and t = 20 mm movie of the base case Inlet temperature, pressure, mass flow rate, and power input measurements of case 2 sampled at 1000 Hz and the mean of the 1000 samples recorded every seven seconds Time-averaged void fractions in the k2ar and k2br regions determined for each of the ten movies associated with the base case and with case Time-averaged void fractions in the k3br and k3cr regions determined for each of the ten movies associated with the base case and case Comparison of the instantaneous void fraction variations in tree 2 during the base case, t = 28 mm movie, and the case 1, t = 31 mm movie. (a) k = 2 level (b) k 3 level Comparison of the instantaneous void fraction in tree 2 during the base case, t = 14 mm movie, and the case 1, t = 25 mill movie. (a) k = 2 level (b) k = 3 level Close-up of the instantaneous void fraction in tree 2 during the base case t = 14 mm movie and case 1 t 25 mm movie. (a) Case 2 (b) Base Case Regional, time series void fraction in the k = 3 level of tree 2 associated with t = 12 mm movie of the base case and t = 10 mill movie ofcase Bifurcations outlined in blue were the focus of the local study Time-averaged void fraction in bifurcation downstream of the k2br channel during the base case... 77

12 Fieure LIST OF FIGURES (Continued) Page 6.34 Variations in local void fraction during t = 3 mill at the bifurcation located downstream of the k2br channel. (a) Full movie (b) t = tot=3.75s Variations in local void fraction during t = 14 mm at the bifurcation located downstream of the k2br channel. (a) Full movie (b) t = 2.21 s tot=2.32s Variations in local void fraction during t = 28 mm at the bifurcation located downstream of the k2bl channel. (a) Full movie (b) t = to t = s (c) t = s to t = s Variations in local void fraction during t = 26 mm at the bifurcation located downstream of the k2bl channel. (a) Full movie (b) Close-up Time-averaged void fraction in bifurcation downstream of the k2br channel during case Variations in local void fraction during t = 4 mm movie of case 1 at the bifurcation located downstream of the k2br channel. (a) Full movie (b) Close-up Time-averaged void fraction in bifurcation downstream of the k2br channel during case Variations in local void fraction during t = 18 mm movie of case 2 at the bifurcation located downstream of the k2br channel. (a) Full movie (b) Close-up... 86

13 Table LIST OF TABLES Page 3.1 Nominal channel dimensions Test matrix... 23

14 NOMENCLATURE C: specific heat (kj/kg K) hg: vapor enthalpy (kj/kg) hf: fluid enthalpy (kj/kg) I: current (A) xh: mass flux (g/min) m1: liquid mass (kg) m: vapor mass (kg) n: number of branches resulting from a bifurcation NWPdata: number of white pixels in data image NWPbe: number of white pixels in base image Q: heat flux (W/cm2) R: resistance (2) Tsat: saturation temperature ( C) T: inlet temperature ( C) Vf: volume of liquid (m3) Vg: volume of vapor (m3) Vj: voltage associated with current measurement (V) V: voltage associated with voltage measurement (V) W: channel width (jim) Xe: exit quality a: apparent void fraction a0i. apparent void fraction

15 Void Fraction Variations in a Fractal-Like Branching Channel Network 1. BACKGROUND In 2005, Motorola unveiled the Razr. This cell phone is smaller than a deck of cards, weighs 95 grams, and has every possible feature, and is a prime example of the current trend towards smaller, more powerful electronics. Powerful electronic devices run hot, making effective cooling crucial. As the size of these devices decrease, the heatsinks used to cool them must also shrink. Heatsinks where heat transfer is the result of fluid flow through microscale channels are one solution currently being explored. The small diameters of the channels allow a large channel density to be achieved and yield larger heat transfer coefficients thereby, maximizing heat transfer while minimizing volume. A clear understanding offlow characteristics in these devices is necessary to exploit their potential. For a given geometry, the only way to improve the heat transfer of singlephase flow is to increase the flow rate. Higher flow rates demand greater pumping power and the benefit-to-cost ratio, the energy transfer rate divided by the flow power, is decreased. Two-phase flow combines the sensible heating of singlephase flow with latent heating. The combination of these two thermodynamic mechanisms allows a given heat transfer rate to be achieved at a much lower flow rate than would be required if the flow was single phase. Extensive studies of twophase, macroscale flow have been performed and the findings used to maximize the benefits while limitations such as increased pressure drops, critical heat flux conditions, or flow instabilities are avoided.

16 Decreasing channel diameter while holding all other conditions fixed, increases flow velocity and decreases the ratio of flow volume to channel surface area. Given a constant Nusselt number, the decrease in diameter associated with moving from a macroscale channel to a microchannel, improves the heat transfer coefficient. Microscale channels also allow a higher channel density than realizable in a macroscale heatsink, increasing the convective surface area per unit volume. Despite these benefits, eventually increasing single-phase microscale heat transfer requires an increase in flow rate. As with macroscale flow, flow boiling offers an alternative. Two-phase flow in microchannels can achieve the same heat transfer rate as single-phase flow at much lower flow rates. In addition, greater temperature uniformity is expected due to the latent heating. Unfortunately, several of the limitations of macroscale two-phase flow have been observed in these microscale configurations as well. Oscillations between single and two-phase flow have been documented, as have oscillations between different flow regimes [1,2,3,4]. Finally, channels that share an inlet plenum often experience flow reversal [1,4]. A perfectly symmetrical, fractal-like branching microchannel geometry has been suggested as an alternative to the naturally asymmetry of a plenum in a parallel channel heatsink [13]. Self-similarity is apparent in the suggested geometry, shown in Fig The term fractal-like is used, because unlike true fractals, the repeated pattern does not continue on to infinity. A key characteristic of this particular design, and inherent to many fractals, is branching. Branching from one large diameter channel into two smaller channels results in an increased flow area. An additional benefit associated with the decreased diameter at each branching level is improved temperature uniformity. The primary drawback of the fractal-like concept is an inability to achieve the same channel density as a

17 3 parallel channel configuration. Two-phase flow in this geometry has not yet been studied and will be the focus of this work. FIGURE 1.1. Fractal-like branching channel network. There are a few terms specific to this design that should be defined. As can be seen in Fig. 1.1, sixteen channels originate from the inlet plenum. Channel depth in the heat sink studied in this analysis remains constant. Following one of these sixteen channels from the inlet plenum outward, we see that the channel splits into two channels of smaller width and length. The split is referred to as a bifurcation. There are five levels associated with this design as shown in Fig The distance from the first bifurcation to the second is a branching level, there are four such levels. The k = 0 level originates from the inlet plenum rather than from another branch; therefore it is not considered a branching level. The design is split into sixteen even sections, a section contains one k is referred to as a tree. Figure 1.2 is an example of a tree. 0 level channel and The objective of this work is to study flow boiling in this fractal-like geometry using high-speed, high-resolution imaging. Flow regimes will be documented and data evaluated for both regional and local void fraction variations. In addition, the effects of this novel geometry will be explored. The impact of bifurcations

18 =4 FIGURE 1.2. Tree of a fractal-like branching channel network. on void fraction, if any, will be studied. The results will be also be examined for interactions between the different branching levels of the geometry. These results will be used to validate a one-dimensional algorithm to be used for optimizing flow geometries for specific operating conditions. Electronic devices are growing smaller and more powerful. Current cooling capabilities are no longer effective, and increased heat transfer capabilities with minimal size has become crucial. Two-phase flow in microscale heatsinks has shown tremendous potential for fulfilling this need. Use of fractal-like heat sinks offer the potential for significant benefit-to-cost ratios. Other potential applications of the these fractal-like networks include use as desorbers or heat exchangers. A complete understanding of flow boiling in microscale channels must be developed in order to maximize this potential. The focus of this study is to develop an understanding of two-phase flow in a fractal-like branching microchannel geometry.

19 5 2. LITERATURE REVIEW AND OBJECTIVES 2.1. Literature Review Microscale Single Phase Flow More than twenty years ago Tuckerman and Pease [6] demonstrated the enhanced heat transfer capabilities of a microchannel heatsink. A potential application of this technology is cooling compact, powerful electronic devices. The electrical properties governing the performance of these systems are temperature sensitive, making uniform cooling a necessity. Typical microchannel heatsinks are composed of constant cross-section channels that are subjected to significant axial heating. Bau [7] demonstrated that tapering the cross-sectional area of a microchannel improves temperature uniformity. Unfortunately a trade-off exists, tapering the channels improves surface temperature uniformity but the diminishing cross-sectional area increases the required pumping power. Mammalian circulatory systems efficiently pump blood throughout the entire body. West et al. [8] proposed that these systems are fractal-like branching networks and developed scaling laws to describe the length and diameter ratios between branching levels. A parallel between metabolic and thermal transport processes was drawn by Pence [9]. By applying the scaling laws of West et al. [8] to a microchannel heatsink design, a fractal-like, branching channel heatsink was generated. The one-dimensional model developed by Pence [10] was used to compare a parallel, straight channel flow network and a fractal-like, branching channel flow network with the same flow power, convective surface area, heat flux, overall channel length, and exit channel width and depth. The fractal-like device experienced pressure drops 35% smaller than the straight channel array. At a

20 bifurcation in the fractal-like design, a single channel branches into two of lesser width. Despite the decrease in hydraulic diameter, the overall flow area is increased, explaining the improved pressure drop performance. Alharbi et al. [11] validated the one-dimensional model results [10] with experimental data and a three-dimensional computational fluid dynamics study. The fractal-like design used in this study is based on fixed length and width ratios, Eq. 2.1 and Eq. 2.2, where the number of daughter channels that result from a mother channel dictates the value of n. As discussed in Chapter 1, the subscript k refers to a lower order branching level while subscript k+ 1 represents the higher order branching level at a given bifurcation. The use of a fixed width ratio differs from the original proposal [8] which called for fixed length and diameter ratios. Limitations in manufacturing processes require the fractal-like channels be of uniform depth. This constraint results in lower branching level channels of infinite width when a fixed hydraulic diameter ratio is employed. A fixed width ratio was substituted for the diameter ratio. An optimization method similar to that used by Bejan [12] was used to define that ratio [13]. Lk+l _1 2 (2.1) Lk Wk+l _1 =n (2.2) Wk 2 n=2 (2.3) Single phase flow studies have shown that the limitations associated with a parallel channel configuration, specifically large pressure drop and surface ternperature non-uniformity, do not occur or occur to a lesser degree in the fractal-like device. These studies need to be extended to two-phase flow.

21 Macroscale Two-Phase Flow In terms of heat transfer, two-phase flow is an attractive alternative to single-phase flow. Unfortunately, it is not without limitations, including the possibility of unstable flow. Instabilities in heat exchanger flow can result in pressure fluctuations and dry-out, greatly affecting overall performance. Small fluctuations in the flow can occur during phase change. The flow can either return to its normal operating conditions or these fluctuations can result in instabilities (Boure et al. [14]). Saha and Zuber [15] evaluated the effects of mass flow rate, heat flux, and system pressure on the onset of significant void (OSV), an indicator of onset of flow instability (OFT). Onset of significant void was dictated by thermal conditions at low mass flow rate conditions. Kennedy et al. [16] studied two-phase flow of deionized water in a single, macroscale channel. Effects of varying mass flow rate and heat flux on the OSV were evaluated, and an equation was developed to predict the heat flux at which OSV, and therefore OFT, would occur. In addition, flow reversal was observed in two-phase water flow in a rectangular channel of hydraulic diameter equal to 0.75 mm by Warner and Dhir [17]. These studies and others like them provided the understanding of two-phase flow necessary to maximize heat transfer in macroscale heat exchangers while avoiding potential issues such as flow instabilities Microscale Two-Phase Flow Two potential issues associated with two-phase flow in microchannels, large pressure drops associated with the phase change and instabilities similar to those seen on the macroscale, have been observed in preliminary studies. Nucleate boiling of a subcooled refrigerant in a parallel channel device resulted in a pressure

22 change one thousand times larger than that measured for single phase flow in the system [18]. Koo et al. [19] saw the pressure drop in a single microchannel increase by 360% after the onset of boiling. Pressure increases of these magnitudes would negate the benefits associated with two-phase flow. As was mentioned earlier, onset of significant void is indicative of an instability. Therefore, quantifying void fraction variations in microchannels is of importance. Void fraction in a horizontal pipe is defined by Wilkes [20] as the portion of the total pipe volume occupied by the gas phase. Tong and Tang [21] assume one-dimensional flow and report void fraction averaged over the crosssectional channel area as a ratio of the gas area to cross-sectional flow area. The first definition approaches void fraction from a volumetric standpoint, while the other uses a ratio of areas. Expressing void fraction in terms of areas is of interest because although several void fraction measurement methods exist, the most common, and the one employed by this study, relies on two-dimensional images. The techniques employed when using images to quantify void fraction were reviewed to determine the requirements for measuring a three-dimensional quantity with a two-dimensional method. Triplett et al. [22] performed a two-phase flow void fraction study in circular and semi-triangular microchannels. The working fluid was an air and water mixture, and the flow was assumed to be onedimensional and steady state. To quantify the void fraction, each bubble was assigned a geometric shape that best represented the actual shape. The volume of the bubble was calculated using the assigned shape and visible dimensions. These measurements compared well to homogeneous flow model predictions. Chung et al. [23] imaged two-phase flow of nitrogen and water in mini and microchannels. In the minichannel, the two-dimensional shape of the bubble was related to a symmetrical volume and the fraction of gas estimated. In the microchannel, it

23 was assumed there was no bubbly flow and all images for a given experiment were assigned a void fraction. All liquid flow was considered to have a void fraction of zero, while annular flow with a very thin liquid film was considered to have a void fraction of unity. Void fraction of annular flow with a thick liquid film is determined by assuming the gas is a cylinder, and taking the ratio of the gas cylinder radius squared to the channel radius squared. The void fractions for each image were summed and divided by the total number of pictures to determine a time averaged void fraction. By quantifying void fraction in a manner similar to the methods just discussed, the results can be evaluated for onset of significant void and, therefore, onset of flow instability. Of the microscale liquid/vapor flow imaging studies reviewed, all reported instabilities. Steinke and Kandlikar [1] investigated two-phase flow of deionized water in a heatsink with six parallel microchannels. The channels were rectangular in cross-section with hydraulic diameters of 207 jim. The flow rate was held constant while the heat flux was increased incrementally and images recorded at 500 frames per second (fps). The flow oscillated between bubbly and annular and counter-flow due to pressure fluctuations, and communication between the parallel channels were observed. The interaction of surface tension, inertia, and evaporation momentum forces resulted in this flow reversal (Kandlikar [24]). Jiang and Wong [2] performed a two-phase flow visualization study of deionized water in two different microchannel heatsinks. Both devices had multipie parallel channels, with hydraulic diameters ranging from 26 to 35 jim. Varying the flow rates and rate of energy input, the authors documented unstable transition between bubbly and annular flow. Zhang et al. [3] also performed a flow visualization study of deionized water flow boiling in a parallel channel network and a single channel device. Flow oscillated between single and two-phase flow, a

24 10 phenomenon attributed to pressure fluctuations. Wu and Cheng [4] investigated flow boiling of deionized water in two different microchannel heatsinks. Flow visualization showed oscillations between single and two-phase flow. Corresponding large amplitude, long period fluctuations in temperature and pressure measurements were recorded. Qu and Mudawar [5] reported hydrodynamic instabilities in two-phase water flow through a parallel channel heatsink. Flow reversal traveling to the inlet plenum was recorded. Pressure drop oscillations were observed for flow that was not throttled upstream of the test device. When the upstream flow was throttled, mild parallel channel oscillations were documented. A review of two-phase microchannel flow studies like those discussed above, lead Bergles and Kandlikar [25] to conclude that two instabilities exist, upstream compressible volume and excursion instabilities. Throttling the flow at the inlet of each channel was proposed to eliminate these instabilities. The heat transfer capabilities of two-phase flow are superior to that of single phase flow and the latent heating associated with the phase change can provide better temperature uniformity along the flow path. Void fractions are an inherent part of two-phase flow and are used to better understand flow boiling. It is for this reason that quantifying void fraction variations is an objective of this study Research Objectives The focus of this parametric study is two-phase flow in the fractal-like branching geometry proposed by Pence and Enfleld [13]. Parameters include inlet mass flow rate and power input. For a given heat flux, the mass flow rate is varied while inlet temperature is held constant.

25 11 High-speed, high-resolution imaging was used to document flow regimes. The images will be evaluated for statistical local and regional void fraction variations. Effects of the bifurcations on vapor flow will be evaluated and interactions between branching levels investigated. Results are to be used to validate the pressure drop predictions of a one-dimensional two-phase algorithm currently being developed at Oregon State University.

26 12 3. TEST DEVICE DESIGN AND MANUFACTURING 3.1. Test Device Design Figure 3.1 provides a cross-sectional view of an ideal test piece. The device is composed of two disks, a bottom layer with the channel network chemically etched and a transparent cover. The working fluid enters the center of the device, and flows radially to exits located along the outer circumference. A constant heat flux applied to the bottom of the test device supplies energy to the flow. The high thermal resistance of the device top ideally prevents energy loss. Glass Silicon disk1 Energy input Energy input out FIGURE 3.1. Cross-sectional view of an ideal test device. An electric heater provides energy input to the flow. A uniform layer of resistive metal deposited on the bottom surface of the bottom disk of the test device forms the heater. A bottom disk with a large thermal conductivity theoretically should provide the desired maximum constant heat flux to the flow. In addition to having a thermally conductive bottom disk, the top disk ideally would be a perfect insulator. The fluid in the test device is defined as the control volume for a global energy balance. A combination of sensible and latent heating of the flow increases the overall heat transfer rate, Q.

27 13 Q iñicp(tsat T) + thixe(hg hf) (3.1) The quality at the test device exit, xe, is defined as Xe m m + m1 (3.2) Boundary conditions are based on the ideal test piece. The glass top is assumed a perfect insulator. By directly depositing heaters on the bottom of a highly conductive disk, a constant heat flux condition is assumed at this boundary. Flow exiting the test device is discharged into an open, annular plenum, therefore exit condition is assumed to be atmospheric pressure Test Device Manufacturing The test device must allow two-phase flow in a fractal-like branching channel network to be visualized. The actual test device is based on the ideal test device design. The fractal-like pattern employed by this study is, for convenience, circular. The device is composed of two individual disks, one silicon and one Pyrex glass. The disk materials were dictated by the thermal requirements discussed in Section 3.1, silicon has a thermal conductivity of 150 W/m K while glass is 1.5 W/m K, and by manufacturing specifications. The fractal-like pattern was dry reactive ion etched (DRIE) 150 micron deep into a silicon wafer 38 mm in diameter and 500 micron thick. The resultant channel cross-section was not perfectly rectangular as hoped. Due to limitations in the etching process, the top of the channel is wider than the bottom. The channel dimensions are shown in Table 3.1. It should be noted that the reported lengths are based on radial distances and not actual channel lengths, see Fig. 3.2.

28 14 TABLE 3.1. Nominal channel dimensions. k level No Depth (tm) Width (/im) Dh (tim) Length (mm) In addition, the hydraulic diameters are not based on a fixed ratio, but are the result of a fixed width ratio and a manufacturing dictated fixed depth. 0 ) FIGURE 3.2. Lengths are radial distances. A Pyrex glass disk, 38 mm in diameter and 500 micron thick, was anodically bonded to the silicon disk forming the fourth wall of the channels. This optically transparent disk allows flow visualization. It should be noted that during the bonding process the glass was scratched. In addition, the glass and silicon disks were not perfectly aligned leaving small portions of some k = 4 level channels uncovered.

29 15 Following the bonding process, a passivation layer was grown on the nonchannel side of the silicon wafer. A layer of nichrome, 5000 A thick, was then deposited on the disk in the ring-like pattern shown in Fig The resistivity of nichrome is ftcm. The nichrome rings function as heaters when an electrical current is passed through them. An electrical insulator was deposited on top of the rings, but small contact pads were not insulated allowing electrical contact with each ring to be made. This particular pattern allows each of the first four branching levels to be heated individually. gs FIGURE 3.3. Nichrome heaters located on bottom side of test device.

30 16 4. EXPERIMENTAL FACILITY AND TEST PLAN 4.1. Manifold A manifold was designed to act as an interface between the test device and flow loop. The manifold provides inlet pressure and temperature measurements, holds the test device firmly in place without obstructing the view of the flow, supplies a current to the resistance heaters on the backside of the test device, and prevents any leaking at the inlet or outer radius of the disk. The manifold body, shown in Fig. 4.1, was machined from polyetheretherketone (PEEK). PEEK is an electrical and thermal insulator with excellent tensile strength at higher temperatures. Inlet pressure and temperature ports were machined in the manifold, 6.96 cm from the test device inlet. The large thermal resistance of the PEEK minimized thermal losses between the inlet measurements and the actual inlet to the test piece. To hold the test device securely in place without obstructing the view of the flow or breaking the delicate test device, the manifold was designed to function as a vacuum chuck. A vacuum pulled through four holes, 1.6 mm in diameter, machined equidistant apart in the manifold top surface, holds the bottom surface of the test device securely against two silicon 0-rings, resting in 0-ring grooves also machined in the surface. The 0-rings, located near the inner and outer radii of the test device, prevent leaking. An additional set of holes, 0.94 mm in diameter, were machined into the top surface of the manifold and gold pogo pins were press fit into them. The contact pads of the nichrome heaters located on the bottom of the test device make electrical contact with the pogo pins when the vacuum is applied. A current is supplied to each heater via a pair of pogo pins.

31 17 body p FIGURE 4.1. Manifold Flow Loop and Instrumentation Flow Loop A flow ioop, shown in Fig. 4.2, provides a preheated working fluid to the test device. The working fluid is heated from room temperature to 80 C in a Vanguard 2000 W, 6 gallon water heater. The warmed flow is passed through a Shelco 10 jim filter and driven at a constant flow rate through the flow loop by a Micropump gear pump capable of 150 g/min at 75 psi. The pump is controlled with a manufacturer supplied rheostat. Two needle valves and a bypass line provide secondary flow control if necessary. The tubing is Swagelok one-quarter inch outer diameter, copper or stainless steel. Just before the test device, onequarter inch, teflon tubing was used to minimize thermal losses and vibrations. This is a closed flow loop, beginning and ending at the water heater. Heating of the flow occurs in stages. The working fluid exits the water heater at 80 C, a temperature that is maintained by an Omega iseries controller with an accu-

32 113 racy of 0.02 C and a T-type thermocouple. Additional energy is added to the flow in a Neslab 800 W constant temperature bath ensuring the fluid leaves at the desired inlet temperature. The flow travels through another 0.6 m of tubing before entering the test device, and although the tubing is well insulated, losses do occur. Therefore, the inlet temperature is fine-tuned using a high temperature, 100 W, McMaster high temperature rope heater located just upstream of the test device inlet. The power output of the rope heater is controlled with a variac and adjusted manually to maintain a constant inlet temperature, measured at the manifold temperature port. Additional energy was added to the flow by resistance heaters on the bottom of the test device. DC constant voltage power supplies powered the heaters. Upstream of the test device a Micromotion Coriolis mass flow meter, with a flow range of 0 to 1370 g/min and an accuracy equal to 0.10% of the rate, measures the inlet mass flow rate. Temperature and pressure ports located in the test manifold allow the inlet temperature and pressure to be measured 6.96 cm from the test device inlet. The temperature is measured with a Therm-X, metal sheathed, 4 wire, resistance temperature detector (RTD) with an accuracy of 0.7 C. A Cole-Parmer capacitance pressure transducer is used to measure the inlet pressure with an accuracy of psig. Liquid and vapor expel from the test device, and the liquid phase is captured in an annular plenum. A vacuum is applied to the plenum to remove vapor leaving the test device. The plenum was open to atmosphere, so the vacuum pulled was minimal, but necessary to prevent fogging of the imaging equipment. Downstream of the plenum, the mass flow rate of the working fluid is measured using another Micromotion Coriolis mass flow meter prior to being returned to the water heater.

33 19 Filter fl Mass flow meter Constant teniperatlue bath Preheater Pump Heaters n atel heater Test Vacuum Mass flow device meter acusnu FIGURE 4.2. Flow ioop Flow Visualization Two-phase flow in the fractal-like branching channel network is recorded with a high-speed, high-resolution imaging system. The imaging set-up is shown in Fig A Nikon 60 mm f/2.8d AF Micro-Nikkor lens was used with a Vision Research Phantom v5.0 camera. The Phantom v5.0 recording speeds range from 60,000 fps at a resolution of 32 x 256 pixels to 1000 fps at 1024 x 1024 pixels. For this study 1000 fps adequately captured void fraction variations and provided a resolution of 630 pixels per cm. An exposure time of 10 jis was used. A Thor three-dimensional translation stage with 1.27 cm travel was used to adjust the camera position. Rhodamine 6G dye was added to deionized, degassed water to create a solution with a molarity of io. The degassing procedure was accomplished by boiling the water for a thirty minute period in the water tank. A constant wavelength Coherent laser with wavelength of 532 nm was used to fluoresce the

34 20 working fluid upon entering the test device. Rhodamine has a peak absorption at 532 urn wavelength laser light and peak emission at 550 nm, as shown in Fig A 570 nm long pass filter allows 60% of the emitted wavelength light to be imaged by the camera. Mirro/. flca Lens JE" Filter Laser Test device 'N Piano convex FIGURE 4.3. Flow visualization loop Data Acquisition Global measurements such as inlet pressure and temperature were output as voltages and sampled at 1000 samples per second using two 16 bit data acquisition (DAQ) boards. Every seven seconds, the mean of the 1000 data samples was calculated and recorded using LabVIEW software. A circuit was built to translate the current and voltage supplied to the resistance heaters to voltages that could be sampled by the DAQ at 1000 samples per second. The inlet mass flow rate was output as a frequency and recorded using Lab VIE W.

35 21 C 0 U) ) Waveiength (nm) FIGURE 4.4. Rhodamine 6G absorption and fluorescence emission spectra from The Phantom v5.o camera was controlled using the manufacturer supplied software. Images were stored in the volatile dynamic ram of the camera. Upon filling the camera memory, 1 GB, the images were extracted using the Phantom software and recorded in Cineon image file format for later conversion to bitmap format and image processing. The image data collection was not synchronized with the global data collection Test Plan The test plan is based on the findings of related works and a battery of preliminary tests. A flow boiling investigation performed in a test device of similar composition to this study's test device with microchannels etched in a silicon bottom with an anodically bonded glass piece forming the top, showed that at higher flow rates the increase in pressure associated with phase change

36 22 broke the silicon to glass bond [3}. These results were used to establish inlet mass flow rate and inlet pressure limits. fluid: Recall that quality is the ratio of the vapor mass to the total mass of the m x = m + mi (4.1) For even a low quality, because the density of liquid water is three orders of magnitude larger than that of water vapor, the vapor volume is large compared to the liquid volume. A sudden, significant change in vapor volume could break the test device, or produce dry out conditions. One the other hand, for this experiment to be successful, boiling must be visible. This goal must be balanced with the need to minimize vapor volume, placing additional constraints on the test plan. Preliminary results also imposed limits on the rate of energy input and dictated a test procedure. A maximum energy input rate of sixty-six watts was set to prevent critical heat flux conditions and/or test device failure. Sixty-one watts represented the minimum energy input that would produce two-phase flow. In addition, it was determined it was most effective to start at a given heat flux and the fastest flow rate, and then incrementally decrease the flow rate. By decreasing the mass flow rate for a given energy input, the amount of boiling was increased slowly. This methodology ensured that if dry-out or test device breakage were to occur it would be at the end of an experiment rather than the beginning. The resultant test matrix is shown in Table 4.1. The required camera frame rate and movie length were dictated by analysis of initial results. Preliminary movies taken at a frame rate of 1000 fps and shutter speed of 10 microseconds allowed the movement of bubbles and the vapor/liquid

37 23 TABLE 4.1. Test matrix. 61W 66W 45 g/min X (Case 1) X (Base Case) 50 g/min X X (Case 2) 55 g/min X X 60 g/min X X 65 g/min X X interface to be clearly visualized. A geometry of 1024 x 1024 pixels allowed one quarter of the test device to be imaged. Movies recorded using this geometry were one second in length due to camera memory limitation of 1 GB. The time period of a one second movie was not long enough to draw conclusions about the flow characteristics. Decreasing the geometry to 512 x 512 pixels, flow from the inlet to the exit in more than one tree was visible and the movie length was increased to seconds increasing the duration to four times that of the original Procedure The same procedure was followed for every test to ensure consistent results. Prior to each experiment the empty water heater was filled with seven liters of deionized water. The water was vigorously boiled for a thirty minute period. After this degassing process, the remaining volume of water was measured and the appropriate amount of Rhodamine 6G added to the water to achieve a molarity of io. A sample containing both rhodamine and water was heated until it began

38 to boil, the boiling point was measured and shown to be the same was that of water. The first step of each experiment was to warm the flow loop. This process began by heating the fluid in the water tank to 80 C. At the same time, the constant temperature bath was brought to 90 C. When these tasks were complete, the warmed fluid was pumped through the system at 100 g/min, bypassing the manifold and test device. The mass flow meters, the largest heat sinks in the ioop, were warmed using heat blankets rather than waiting the hours required for the flow to bring them to temperature. When the fluid passing through the mass flow meters reached 70 C, the warm-up procedure was complete. The desired flow rate was set and flow sent to the manifold and test device. At this point, the camera was positioned and the test device illuminated. Images of the fractal full of liquid were recorded. These base images were required for data analysis, which are discussed in Chapter 5. The inlet temperature of the fluid was then increased to the desired level by adjusting the power output of the rope heater located just upstream of the inlet. There is an order of magnitude difference between the coefficient of thermal expansion of silicon and that of glass; therefore, the test device heat flux was incrementally increased to the desired level. When the inlet mass flow rate of two-phase flow and the energy input were constant for thirty minutes, the experiment was considered steady-state. Once the experiment was steady-state, ten movies were recorded over the next thirty minutes. Each movie had an image size of 512 x 512 pixels, a shutter speed of 10 microseconds, and a frame rate of 1000 fps. Transferring a movie from the camera memory to the computer took approximately three minutes. The time period associated with this transfer and an experiment length of thirty minutes dictated that ten movies could be recorded per experiment. 24

39 After ten movies for a particular flow rate and power were recorded, the flow rate was decreased by 5 g/min and the imaging procedure repeated. The flow rate was decreased a total of four times while the power was held constant. After the fifth experiment was finished, the shut down procedure was implemented. The power was then decreased incrementally, again to avoid thermal mismatch, until the test device was no longer heated. At this point, all heaters were turned off and the flow gradually slowed to a stop. The test device was flushed with deionized water to prevent the Rhodamine 6g from staining the silicon and the flow loop was shut down. 25

40 5. DATA REDUCTION AND ANALYSIS 5.1. Data Reduction Global Data Reduction Global measurements of pressure, temperature, mass flow rate, and energy input quantify the inlet conditions of each experiment. All inlet measurements were made for the duration of the steady state experiment. The instrument outputs were sampled at 1000 samples per second, and the mean of the 1000 data samples was calculated. The mean was recorded, writing the data to an output file took 7 seconds, after which another 1000 samples were measured and the process repeated. The voltage output of the pressure transducer and RTD were converted to pressures and temperatures using the calibration equations shown in APPENDIX A. The inlet mass flow rate measurements were converted from frequencies to flow rates using a linear relationship, APPENDIX A. A circuit translated the currents and voltages supplied to the resistance heaters into voltages, V1 and V, respectively, that could be measured by the DAQ. The voltage drop measured across a current sense resistor was used to calculate the current, I, supplied to the heater, Eq The voltage supplied to the heater was too large to be read by the DAQ and was sent through a voltage divider before being measured. The actual voltage supplied to the heater, V, was calculated using Eq The calculated current and voltage values were used to determine the nominal rate of heat transfer. Once corrected for thermal losses, qiosses, the actual heat transfer rate, q, is calculated using Eq Thermal losses include conduction from the test device bottom to the manifold and radiation and convection from the test device top. Conduction losses from the test device

41 27 bottom to the manifold were neglected because the thermal conductivity of PEEK, the manifold material, is 160 times smaller than that of silicon. Losses due to convection and radiation were estimated to be 2 W, about 3% of the maximum energy input. VI Rsense (5.1) R1UMO V=V* (5.2) R1OMO + R470k0 q = (V * I) qiosses (5.3) A constant heat flux boundary condition was assumed at the bottom of the test device. Several possible heat transfer areas exist. The area of the device bottom, where the heaters are located, is 11.4 cm2. This area was used to determine the heat flux applied to the flow. It should be noted that, the four heater rings deposited on the bottom of the test device cover 8.3 cm2 of the available 11.4 cm2. An 0-ring at the outer diameter of the test device prevents the fourth heater ring from being used, decreasing the heater area to 6.15 cm2. In addition, the convective surface area is 5 cm2. Two other factors affecting the constant heat flux boundary should be noted. The k = 4 level does not have a heater ring directly below it, as the other branching levels do, and relies on conduction for its energy input. An o-ring at the inlet of the test device causes pooling of hot water in the area between the 0-ring and the inlet plenum potentially increasing the temperature of the disk in this area. Despite these issues, because of the high thermal conductivity of the silicon, 150 W/m K, the constant heat flux boundary is considered a safe assumption.